Scheduling disruptive work.
WHAT NEEDS A WINDOW
Version upgrades Large schema changes Hardware or storage changes Failover testing
WHAT TO ESTABLISH
How long it will actually take.
HOW
By performing it on a copy, timed.
WHY THAT IS ESSENTIAL
Estimates for database operations are frequently wrong by an order of magnitude.
WHEN TO SCHEDULE
When query volume is lowest.
HOW TO KNOW
Query rate by hour, recorded over a week.
WHAT TO COMMUNICATE
The window, the expected impact, and to whom.
WHAT TO PREPARE
The exact steps, written A verification checklist A way back A decision point for abandoning
WHY A DECISION POINT
Without one, people continue past the window hoping.
WHAT TO DO BEFORE STARTING
A backup, verified.
WHAT TO DO DURING
Follow the written steps Record what actually happens
WHAT TO VERIFY BEFORE DECLARING COMPLETION
Applications connect Key queries return expected results Replication is running Performance is unchanged
WHAT TO DO IF IT OVERRUNS
Decide at the point you set, rather than drifting.
WHAT TO RECORD AFTERWARDS
Actual duration, for the next time.
